✏Encrypted Client Hello (ECH) in Censorship Circumvention
Free and Open Communications on the Internet (FOCI), 2025
Publications
Here you can find my publications ordered by publication date. First author papers are marked with ✏.
Conference Papers
✏Transport Layer Obscurity: Circumventing SNI Censorship on the TLS-Layer (Distinguished Paper Award)
IEEE Symposium on Security and Privacy (SP), 2025
Abstract
HTTPS composes large parts of today’s Internet traffic and has long been subject to censorship efforts in different countries. While censors analyze the Transport Layer Security (TLS) protocol to block encrypted HTTP traffic, censorship circumvention efforts have primarily focused on other protocols such as TCP. In this paper, we hypothesize that the TLS protocol offers previously unseen opportunities for censorship circumvention techniques. We tested our hypothesis by proposing possible censorship circumvention techniques that act on the TLS protocol. To validate the effectiveness of these techniques, we evaluate their acceptance by popular TLS servers and successfully demonstrate that these techniques can circumvent censors in China and Iran. In our evaluations, we discovered 38—partially standard-compliant—distinct censorship circumvention techniques, which we could group into 11 unique categories. Additionally, we provide novel insights into how China censors TLS traffic by presenting evidence of at least three distinct censorship appliances. We suspect that other parts of China’s censorship apparatus and other censors exhibit similar structures and advocate future censorship research to anticipate them. With this work, we hope to aid people affected by censorship and stimulate further research into censorship circumvention using cryptographic protocols.I(ra)nconsistencies: Novel Insights into Iran’s Censorship
Free and Open Communications on the Internet (FOCI), 2025
Abstract
Iran employs one of the most prominent Internet censors in the world. An important part of Iran’s censorship apparatus is its analysis of unencrypted protocols such as HTTP and DNS. During routine evaluations of Iran’s HTTP and DNS censorship, we noticed several properties we believe to be unknown today. For instance, we found injections of correct static IPs for some domains such as google.com on the DNS level, unclear HTTP version parsing, and correlations between DNS and HTTP censorship. In this paper, we present our findings to the community and discuss possible takeaways for affected people and the censorship circumvention community. As some of our findings left us bewildered, we hope to ignite a discussion about Iran’s censorship behavior. We aim to use the discussion of our work to execute a thorough analysis and explanation of Iran’s censorship behavior in the future.TLS-Attacker: A Dynamic Framework for Analyzing TLS Implementations (2nd Place Impact Award)
ACSAC, 2024
Abstract
TLS-Attacker is an open-source framework for analyzing Transport Layer Security (TLS) implementations. The framework allows users to specify custom protocol flows and provides modification hooks to manipulate message contents. Since its initial publication in 2016 by Juraj Somorovsky, TLS-Attacker has been used in numerous studies published at well-established conferences and helped to identify vulnerabilities in well-known open-source TLS libraries. To enable automated analyses, TLS-Attacker has grown into a suite of projects, each designed as a building block that can be applied to facilitate various analysis methodologies. The framework still undergoes continuous improvements with feature extensions, such as DTLS 1.3 or the addition of new dialects such as QUIC, to continue its effectiveness and relevancy as a security analysis framework.Turning Attacks into Advantages: Evading HTTP Censorship with HTTP Request Smuggling
Free and Open Communications on the Internet (FOCI), 2024
Abstract
Many countries limit their residents' access to various websites. As a substantial number of these websites do not support TLS encryption, censorship of unencrypted HTTP requests remains prevalent. Accordingly, circumvention techniques can and have been found for the HTTP protocol. In this paper, we infer novel circumvention techniques on the HTTP layer from a web security vulnerability by utilizing HTTP request smuggling (HRS). To demonstrate the viability of our techniques, we collected various test vectors from previous work about HRS and evaluated them on popular web servers and censors in China, Russia, and Iran. Our findings show that HRS can be successfully employed as a censorship circumvention technique against multiple censors and web servers. We also discover a standard-compliant circumvention technique in Russia, unusually inconsistent censorship in China, and an implementation bug in Iran. The results of this work imply that censorship circumvention techniques can successfully be constructed from existing vulnerabilities. We conjecture that this implication provides insights to the censorship circumvention community beyond the viability of specific techniques presented in this work.✏Poster: Circumventing the GFW with TLS Record Fragmentation
ACM CCS, 2023